The purpose of this project was to develop a method for producing confidence intervals for the attributable risk in case-control studies when exposure occurs at multiple levels and to demonstrate that the method has satisfactory properties for practical applications.